A DISTRIBUTED EVENT PROCESSING METHOD FOR GENERAL-PURPOSE COMPUTATION

Previous proposals for the application of discrete event oriented meth ods to automatic parallelization have been based on the optimistic exe cution strategy. In this paper we present a new method which avoids op timistic execution. This is motivated by the observation that the cont rol structure of a conventional program constitutes a temporal coordin ate system which is exogenous to the program execution. The method emp loys a logical time mechanism and provides adaptive synchronisation fo r the distributed execution. Hence Data dependent and/or conditional p arallelism is released without the risk of coherency violation. The pa per begins with a brief introduction to Parallel Discrete Event Simula tion (PDES) paradigm. Efficient coarse grain mapping of conventional p rograms onto this paradigm is then discussed. (C) 1998 Elsevier Scienc e B.V. All rights reserved.